Predictors of Death and Vascular Events in the Elderly
The Perth Community Stroke Study
Background and Purpose—The goal of the present study was to identify risk factors for vascular disease in the elderly.
Methods—We conducted a prospective study of control subjects from a population-based study of stroke in Perth, Western Australia, that was completed in 1989 to 1990 and used record linkage and a survey of survivors to identify deaths and nonfatal vascular events. Data validated through reference to medical records were analyzed with the use of Cox proportional hazards models.
Results—Follow-up for the 931 subjects was 88% complete. By June 24, 1994, 198 (24%) of the subjects had died (96 from vascular disease), and there had been 45 nonfatal strokes or myocardial infarctions. The hazard ratio for diabetes exceeded 2.0 for all end points, whereas the consumption of meat >4 times weekly was associated with a reduction in risk of ≤30%. In most models, female sex and consumption of alcohol were associated with reduced risks, whereas previous myocardial infarction was linked to an increase in risk.
Conclusions—There are only limited associations between lifestyle and major vascular illness in old age. Effective health promotion activities in early and middle life may be the key to a longer and healthier old age.
Like many other countries, Australia is facing a rapid expansion in its elderly population due to aging of the postwar cohort of “baby boomers.” As these citizens age, their absolute risk of death and serious disabling nonfatal vascular events increases. Controversy exists among providers of primary and secondary prevention regarding whether to direct health promotion resources and strategies to the young and middle aged or to those at the greatest immediate risk: the elderly. We sought to identify any potentially modifiable lifestyle behaviors that have an independent and significant impact on the risk of death and serious vascular events among an elderly population of Australians.
In a previous, population-based case-control analysis related to the Perth Community Stroke Study (PCSS), we demonstrated that smoking, the addition of salt to food after it was cooked, the frequent consumption of meat, and the use of full-fat milk were all associated with an increased risk of stroke, whereas a regular intake of fish and the consumption of up to 2 “standard” drinks daily (equivalent to 20 g of alcohol) were associated with a lower risk of stroke.1 Because some of these findings, such as that related to the consumption of meat ≥4 times weekly, were novel, we took the opportunity to prospectively follow up the cohort of control subjects during a period of 4 years. Because the control subjects were matched to case subjects for age and because the median age of the case subjects at the time of the onset of their stroke was 75 years, end points have accrued rapidly. In the present report, we therefore consider risk factors for all major vascular events combined, first-ever major vascular events, all strokes, deaths from vascular disease, and deaths from any cause in the elderly.
Subjects and Methods
Recruitment of Cohort
The methods used to identify cases of stroke and control subjects for our earlier case-control study of risk factors for stroke have been described in detail elsewhere.1 Briefly, during a period of 18 months from February 1989 on, the PCSS investigators compiled a register of all cases of suspected stroke or transient cerebral ischemic attack that occurred in a geographically defined population of ≈138 000 persons living in the northern suburbs of Perth, the capital city of the state of Western Australia. For each patient in whom a final diagnosis of stroke was made, we selected between 1 and 5 control subjects from the electoral rolls for the PCSS population, matching for sex and 5-year birth cohort of age. Because enrollment to vote is compulsory for Australian citizens aged ≥18 years and because the response fraction among potential control subjects was 75%, the cohort of control subjects should be reasonably representative of the general (elderly) population. The only small exception is that in a few instances, an individual who was invited to be a control subject was found to have had a stroke during the period of the study that had not previously been registered. These subjects were then included as case subjects, and new control subjects were selected.1
Potential control subjects were contacted by mail and invited to participate in the study. Those who agreed were interviewed at home by 1 of 3 research nurses or by C.S.A. Interviews began early in 1990 and were completed in mid-1991. Independent variables of interest included demographic details such as marital status, history of vascular conditions and diabetes, present functional status, smoking and drinking habits, and a range of dietary practices. Each subject provided written, informed consent to participate in the study and to the use of information they provided for purposes of research.
The identification of information for control subjects was linked electronically to name-identified unit mortality records for the entire state of Western Australia through June 24, 1994, and to the Hospital Morbidity Data System, a name-identified electronic collection of data pertaining to inpatient admissions to hospitals throughout the state that is maintained by the Health Department of Western Australia. All diagnoses of vascular disease apparent from either system, regardless of whether the diagnosis was the principal or a secondary cause of the death or of the admission to hospital, were subsequently followed up with the certifying or attending physician.
The identification of information for control subjects who were apparently alive at the follow-up date was linked to an electronic copy of the state electoral roll to obtain a current address; these individuals were then sent a reply-paid postal questionnaire that inquired about their current state of health and lifestyle and the occurrence of any coronary or cerebrovascular symptoms since the initial interview for the PCSS. If the subject had apparently had a new vascular event, supplementary questions were asked to identify which physician, if any, was consulted and for permission for the study team to contact that physician for further details of the clinical event.
Clinical details of possible vascular events were abstracted from relevant medical records or obtained through interviews with the physicians and were recorded on standardized forms by S.F. This information was subsequently assessed in the light of the standard World Health Organization definitions for acute myocardial infarction (AMI)2 and stroke3 by K.J. or G.J.H., respectively.
The end points of interest in this study included death from any cause, death from vascular disease (including AMI, ischemic heart disease [IHD], cerebrovascular disease, peripheral vascular disease, aortic aneurysm, and mesenteric thrombosis), major vascular events (fatal vascular events and nonfatal AMI or stroke, with only the first such episode occurring during the follow-up of a given individual being taken into account in the analyses), first-ever major vascular events (in subjects who reported no history of AMI or cerebrovascular disease at their original interview), and all strokes. An event in the coronary or cerebrovascular tree was deemed fatal if death occurred within 28 days of the onset of symptoms and new symptoms occurring in the same arterial distribution during that period were not counted as second events. After preliminary bivariate comparisons, the relationships between endpoints and demographic variables and aspects of lifestyle and medical history recorded in the initial interviews with the subjects were studied further with the use of Cox proportional hazards models, with the application of a censor date of June 24, 1994, and the use of reverse stepwise multivariate modeling and EGRET software.4 All models included adjustment for sex and both linear and quadratic terms for age because preliminary analyses indicated significant improvements in fit when the latter was included. A probability value of 5% was regarded as significant in all phases of the analysis, including selection of variables for initial multivariate models.
The protocol for the PCSS and for the present follow-up study was approved by the Committee for Human Rights at the University of Western Australia and by the Confidentiality of Health Information Committee of the Health Department of Western Australia. All subjects provided informed consent.
Completeness of Follow-Up and Number of End Points
Follow-up information was obtained for 817 (88%) of the 931 control subjects originally recruited to the PCSS. By June 24, 1994, 198 (24%) individuals were known to have died, and in 96 of these cases, the single official underlying cause of death was given as a manifestation of vascular disease. An additional 15 and 30 nonfatal episodes of AMI and stroke, respectively, were identified. Overall, between the first interview and mid-1994, there were 141 major vascular events (vascular death, nonfatal myocardial infarction, or stroke) in 126 individuals. During the same period, ≥1 stroke (nonfatal or fatal) occurred in 46 individuals.
Of the 658 individuals who reported no history of coronary or cerebrovascular disease at their initial interview, 83 (13%) had a first-ever vascular event during follow-up.
Death From Any Cause
After adjustment for age and sex, the variables that were predictive of death from any cause in bivariate comparisons included any evidence at entry of established disability (Barthel ADL Index5 of <20, Rankin score6 of ≥3, or Frenchay Activities Index7 of ≤45); a history of myocardial infarction, stroke, or diabetes mellitus; smoking status; frequent consumption of meat; and not being a current drinker of alcohol. There were no significant relationships with consumption of fish, the trimming fat from meat or the skin from chicken, the use of butter as opposed to margarine, the use of full-fat as opposed to reduced-fat or skim milk, the addition of salt to food after it was cooked, a history of claudication or hypertension, being unmarried or widowed, or living alone at baseline.
As shown in Table 1⇓, the final multivariate predictive model for death from any cause omitted whether the subject had a history of stroke and was a current drinker but did indicate a protective effect associated with regular use of aspirin for any reason. Smokers of >20 cigarettes daily at baseline were at 3 times the risk of dying during subsequent follow-up than were lifelong nonsmokers.
Fatal Vascular Events
Again, the 96 fatal vascular events (death from coronary disease, stroke, ruptured aortic aneurysm, peripheral vascular disease, or mesenteric thrombosis) were associated with a number of variables in 2-way comparisons (Table 2⇓). The corresponding multivariate model, which was based on 787 subjects with complete data at baseline, simplified this list to 6 variables, of which only 2 (consumption of alcohol and consumption of meat) were associated with a reduction in risk and were under the direct control of the individual. In contrast to deaths from all causes (Table 1⇑), established disability was not predictive of vascular deaths, and there was no significant association with current or previous smoking habits.
Major Vascular Events
There were 141 major vascular events (death from coronary disease, stroke, ruptured aortic aneurysm, peripheral vascular disease or mesenteric thrombosis, or nonfatal myocardial infarction or nonfatal stroke) in 126 individuals. The final model shown in Table 3⇓ is based on data from 787 subjects and is the only model in the series to indicate a significant protective effect associated with the habitual use of reduced-fat or skim milk. Individuals who consumed alcohol more than “infrequently” at baseline had a significant reduction in risk compared with nondrinkers. Participants who were current smokers (as opposed to ex-smokers and lifelong nonsmokers combined) at entry to the study had a reduced risk of major vascular events, but this was confined to those smoking ≤20 cigarettes daily.
A total of 43 individuals had ≥1 confirmed stroke during follow-up. In a multivariate model with adjustment for age (linear and quadratic terms) and sex, only established diabetes was associated with a significant excess risk of stroke (hazard ratio 3.77, 95% CL 1.71 to 8.31). The addition of a variable indicative of a history of stroke or transient cerebral ischemic attack reduced this point estimate to 3.58 (95% CL 1.62 to 7.91) and was itself of borderline significance (hazard ratio 2.08, 95% CL 0.99 to 4.39).
First-Ever Major Vascular Events
There were 83 first-ever major vascular events (as defined earlier) in the 658 subjects who gave no history of AMI or stroke at the original interview. Even so, bivariate comparisons did not reveal any relationships not already seen when the data for all subjects were examined, whereas the corresponding multivariate model included only 3 factors: diabetes mellitus, Rankin score, and consumption of meat. The reduction in risk evident in women was lesser for this end point than for any of the others considered and did not reach statistical significance in the final model.
Although we previously observed that a range of factors related to lifestyle and behavior were associated with significant variations in the risk of stroke, only a minority of these proved to be consistent predictors of major vascular events, including strokes, when studied prospectively. Apart from age, established diabetes mellitus at the time of initial interview was consistently associated with an increase in risk, whereas the consumption of meat ≥5 times weekly on a regular basis appeared to confer a protective effect. Female sex and some consumption of alcohol were also associated with reductions in risk but did not appear in all 4 final statistical models. Other statistically significant relationships that were consistently apparent in the crude data, such as an increase in risk of 15% to 20% associated with being unmarried or widowed, did not persist after age and sex were taken into account. The relative importance of new potential prognostic factors (eg, risk factors and treatments) that may have arisen after the initial interview and during the follow-up period was not analyzed because this information was not collected.
The validity of the present results should be enhanced by the prospective design of the study, with avoidance of recall bias in relation to the exposures of interest. Although we were able to determine the vital status for at least 88% of the original cohort of control subjects, a large proportion of the end points in the present analyses were fatal illnesses in quite elderly patients and, in many instances, few investigations had been undertaken and the contemporaneous records of symptoms and signs were limited. Thus, despite a careful review of the information available for each event, our data may be subject to considerable diagnostic inaccuracy. Where such errors are random, they will have caused the observed hazard ratios to be biased toward the null, as will any changes in lifestyle and behavior made by individual participants during follow-up. In addition, there may be systematic errors in certification of deaths such as a preference for a final diagnosis of stroke or heart attack in an already infirm elderly patient with a history of vascular disease because it potentially raises fewer questions about testamentary capacity than does a diagnosis of Alzheimer’s disease. Diagnostic preferences of this kind would also tend to obscure true associations between lifestyle and vascular events in the elderly if those associations exist.
The moderate but consistent protective effect associated with relatively frequent consumption of meat was an unexpected finding in this study and one that contradicts our earlier results for myocardial infarction in middle-aged men8 and for stroke.1 Although it is well established that vegetarians have lower mortality rates from all causes than do omnivores,9 the significance of different sources of animal protein in the omnivore diet has received only limited attention. With regard to stroke, the Zutphen Elderly Study10 and the Chicago Western Electric Study,11 respectively, support and provide no evidence for a significant protective effect associated with the regular consumption of fish. On the other hand, the Chicago Western Electric Study found that a high consumption of fish is associated with a significant protective effect for coronary deaths, although a degree of inconsistency concerning such a relationship is also apparent in the literature.12 Non–meat-eaters have a lower coronary mortality rate than do meat-eaters, but this observation may be subject to confounding by other systematic differences in diet and lifestyle between these groups.9 13 For example, the consumption of meat is inversely related to socioeconomic status, as judged by previous occupation, among the elderly men in the Zutphen Elderly Study.14 Unfortunately, we did not collect data on income or occupational history and therefore cannot examine whether the protective effects associated with the frequent consumption of meat in our study are really a reflection of underlying socioeconomic factors. In any case, the modest and inconsistent relationships seen among meat, fish, and vascular disease in the literature overall suggest that any true effects are likely to be small.
The same cannot be said for alcohol, for which a careful review of the evidence demonstrates a J-shaped relationship between consumption and risk.15 Our data that showed lower risks in all drinkers but increased risks, relative to teetotalers, in those consuming >6 standard drinks daily fit well with this conclusion.
Because the median age of our cohort at entry to the study was 75 years, it is conceivable that other modifiable aspects of lifestyle have only a limited impact on the health of the already elderly in the medium term. Even if some individuals had made generally recommended changes to their lifestyle, such as a conscious effort to reduce their intake of saturated fat, in response to campaigns run in Western Australia during the 1980s, the limited duration of our follow-up and the long period over which they had taken less care in such matters may have combined to make it very difficult for us to discern a protective effect flowing from the change.
A striking feature of our data is the consistency with which diabetes mellitus and symptomatic vascular disease predict a poorer outcome in the elderly. Both of these conditions are at least in part preventable. Individuals with established disability as measured with the Barthel ADL Index,5 the Rankin score6 or modified Oxford Handicap Scale to measure a broader need for assistance, or the Frenchay Activities Index7 to measure performance of a variety of roles, such as shopping or visiting friends, also tended to be at an increased risk of further adverse events. Taken together with our finding of only limited direct associations between lifestyle and behavior at entry to the study and major vascular illness in the medium term, it is tempting to conclude that energetic implementation of effective health promotion activities aimed at individuals in early and middle life is likely to be the key to a longer and healthier old age. Given the rapid expansion of the elderly population that is now occurring in many countries, an adequate test of this hypothesis is urgently required to setting an appropriate balance between preventive and treatment services.
This work was supported by Healthway, the Western Australian Health Promotion Foundation. We are grateful to the control subjects and their physicians for continued assistance with the PCSS and to the Health Department of Western Australia, the State Electoral Commission, and the Registrar-General for Births, Deaths and Marriages for Western Australia for access to records within their keeping. Some of the original interviews with control subjects were conducted by Joyce Lim and Jennie Linto. The principal investigators for the PCSS include Dr Ted Stewart-Wynne and Prof Peter Burvill. The protocol for the PCSS and for this follow-up study was approved by the Committee for Human Rights at the University of Western Australia and by the Confidentiality of Health Information Committee of the Health Department of Western Australia.
- Received October 25, 1999.
- Revision received January 24, 2000.
- Accepted January 24, 2000.
- Copyright © 2000 by American Heart Association
Jamrozik K, Broadhurst RJ, Anderson CS, Stewart-Wynne EG. The role of lifestyle factors in the etiology of stroke: a population-based case-control study in Perth, Western Australia. Stroke. 1994;25:51–59.
Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas A-M, Pajak A, for the WHO MONICA Project. Myocardial infarction and coronary deaths in the World Health Organization MONICA Project: registration procedures, event rates and case fatality in 38 populations from 21 countries in 4 continents. Circulation. 1994;90:583–612.
Epidemiological Graphics, Estimation, and Testing Package, Version 0.26.6. Seattle, Wash: SERC; 1991.
Wolfe CD, Taub NA, Woodrow EJ, Burney PG. Assessment of scales of disability and handicap for stroke patients. Stroke. 1991;22:1242–1244.
Wade DT, Legh-Smith J, Langton Hewer R. Social activities after stroke: measurement and natural history using the Frenchay Activities Index. Int Rehab Med. 1985;7:176–181.
Spencer CA, Jamrozik K, Lambert LJ. Do simple prudent health behaviours protect men from myocardial infarction? Int J Epidemiol. 1999;28:846–852.
Snowdon DA. Animal product consumption and mortality because of all causes combined, coronary heart disease, stroke, diabetes, and cancer in Seventh-Day Adventists. Am J Clin Nutr. 1988;48(3 suppl):739–748.
Keli SO, Feskens EJ, Kromhout D. Fish consumption and risk of stroke: the Zutphen Study. Stroke.. 1994;25:328–332.
Orencia AJ, Daviglus ML, Dyer AR, Shekelle RB, Stamler J. Fish consumption and risk of stroke in men: 30-year findings of the Chicago Western Electric Study. Stroke. 1996;27:204–209.
Thorogood M, Mann J, Appleby P, McPherson K. Risk of death from cancer and ischaemic heart disease in meat and nonmeat eaters. BMJ. 1994;308:1667–1670.
Huijbregts PP, Feskens EJ, Kromhout D. Dietary patterns and cardiovascular risk factors in elderly men: the Zutphen Elderly Study. Int J Epidemiol. 1995;24:313–320.
English DR, Holman CDJ, Milne E, Winter MG, Hulse GK, Codde JP, Bower CI, Corti B, de Klerk N, Knuiman MW, Kurinczuk JJ, Lewin GF, Ryan GA. The Quantification of Drug-Caused Morbidity and Mortality in Australia. Canberra, Australia: Commonwealth Department of Human Services and Health; 1995.